2022年被引用次数最多的100篇人工智能论文

Frontier 3月 06, 2023 2.84w 浏览 0

谁正在发表最具影响力的 AI 研究？随着人工智能创新的飞速发展，尽快获取一些前沿信息至关重要。没有人有时间阅读所有内容，但这 100 篇论文肯定会为我们的 AI 技术发展方向指明方向。研发团队影响力的真正考验当然是技术如何出现在产品中，OpenAI 在 2022 年 11 月底发布了 ChatGPT，震惊了世界，紧随其 2022 年 3 月的论文“训练语言模型以遵循人类反馈的指令” ”。如此快速的产品采用很少见，因此为了进一步了解，我们来看一个经典的学术指标：引用次数。通过对 2022 年、2021 年和 2020 年每年被引用次数最多的 100 篇论文进行详细分析，我们可以得出一些早期结论。美国和谷歌仍然占据主导地位，DeepMind取得了卓越的成功，但考虑到OpenAI的产值，它确实是自成一派，在产品影响力和迅速广泛引用的研究方面都如此。

利用Zeta Alpha平台的数据和精心维护（有关方法学的更多信息在下文），平台收集了2022年、2021年和2020年人工智能领域的引用率最高的论文，并分析了作者的附属机构和国家。这使得我们可以按研发影响力而不是纯出版量对其进行排名。找有价值的信息，请记住Byteclicks.com

平台还添加了推特提及，这有时被看作是早期影响的指标，但到目前为止，它们之间的相关性似乎很弱，需要进一步研究。

以下是前100名列表，包括标题、引用次数和附属机构。

Title	Tweets	Citations	Organizations	Countries	Org Types
AlphaFold Protein Structure Database: massively expanding the structural coverage of protein-sequence space with high-accuracy models		1331	European Molecular Biology Laboratory	–	Academia
ColabFold: making protein folding accessible to all		1138	Max Planck Institute for Multidisciplinary Sciences	Germany	Academia
A ConvNet for the 2020s	857	835	Meta, UC Berkeley	USA, USA	Industry, Academia
Hierarchical Text-Conditional Image Generation with CLIP Latents	105	718	OpenAI	USA	Industry
PaLM: Scaling Language Modeling with Pathways	445	426	Google	USA	Industry
Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding	2462	390	Google	USA	Industry
Instant Neural Graphics Primitives with a Multiresolution Hash Encoding	11	342	NVIDIA	USA	Industry
SignalP 6.0 predicts all five types of signal peptides using protein language models		274	Technical University of Denmark, ETH Zurich	Denmark, Switzerland	Academia, Academia
Swin Transformer V2: Scaling Up Capacity and Resolution	87	266	University of Science and Technology of China	China	Academia
Training language models to follow instructions with human feedback	448	254	OpenAI	USA	Industry
Chain of Thought Prompting Elicits Reasoning in Large Language Models	378	224	Google	USA	Industry
Flamingo: a Visual Language Model for Few-Shot Learning	71	218	DeepMind	UK	Industry
Classifier-Free Diffusion Guidance	53	194	Google	USA	Industry
Magnetic control of tokamak plasmas through deep reinforcement learning	0	194	DeepMind	UK	Industry
data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language	0	191	Meta	USA	Industry
OPT: Open Pre-trained Transformer Language Models	812	187	Meta	USA	Industry
BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation	79	184	Salesforce	USA	Industry
A Generalist Agent	231	180	DeepMind	UK	Industry
LaMDA: Language Models for Dialog Applications	473	180	Google	USA	Industry
CMT: Convolutional Neural Networks Meet Vision Transformers	0	172	University of Sydney	Australia	Academia
Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model	271	158	Microsoft	USA	Industry
What Makes Good In-Context Examples for GPT-3?	0	157	Duke University	USA	Academia
Ensemble unsupervised autoencoders and Gaussian mixture model for cyberattack detection		145	Ningbo University of Technology	China	Academia
Training Compute-Optimal Large Language Models		144	DeepMind	UK	Industry
Learning robust perceptive locomotion for quadrupedal robots in the wild	3	141	ETH Zurich	Switzerland	Academia
Do As I Can, Not As I Say: Grounding Language in Robotic Affordances	82	135	Google	USA	Industry
How Do Vision Transformers Work?	193	129	Yonsei University, NAVER	South Korea, South Korea	Academia, Industry
Scaling Up Your Kernels to 31×31: Revisiting Large Kernel Design in CNNs	30	127	Tsinghua University	China	Academia
Large Language Models are Zero-Shot Reasoners	862	124	University of Tokyo	Japan	Academia
Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time	0	122	University of Washington	USA	Academia
Patches Are All You Need?	117	116	Carnegie Mellon University	USA	Academia
Competition-Level Code Generation with AlphaCode		113	DeepMind	UK	Industry
TensoRF: Tensorial Radiance Fields	73	110	ShanghaiTech University	China	Academia
Video Diffusion Models	0	103	Google	USA	Industry
Data Analytics for the Identification of Fake Reviews Using Supervised Learning		102	Dr. Babasaheb Ambedkar Marathwada University	India	Academia
Visual Prompt Tuning	26	102	Cornell University	USA	Academia
DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection	15	100	Hong Kong University of Science and Technology	Hong Kong	Academia
VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training	66	100	Nanjing University, Tencent	China, China	Academia, Industry
Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?	199	99	University of Washington, Meta	USA, USA	Academia, Industry
BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers	11	96	Nanjing University, Shanghai AI Lab	China, China	Academia, Academia
Conditional Prompt Learning for Vision-Language Models	51	93	Nanyang Technological University	Singapore	Academia
Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distribution	151	93	Stanford University	USA	Academia
Measuring and Improving the Use of Graph Information in Graph Neural Networks	1	93	Chinese University of Hong Kong	Hong Kong	Academia
Exploring Plain Vision Transformer Backbones for Object Detection	205	91	Meta	USA	Industry
GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation	26	90	Mila, University of Montreal	Canada, Canada	Academia, Academia
OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework	91	88	Alibaba Group	China	Industry
Block-NeRF: Scalable Large Scene Neural View Synthesis	641	86	UC Berkeley	USA	Academia
Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents	24	86	UC Berkeley	USA	Academia
Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models	881	81	University of Notre Dame	USA	Academia
Outracing champion Gran Turismo drivers with deep reinforcement learning		80	Sony	Japan	Industry
BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning	10	77	Google	USA	Industry
DN-DETR: Accelerate DETR Training by Introducing Query DeNoising	0	74	Hong Kong University of Science and Technology	Hong Kong	Academia
Equivariant Diffusion for Molecule Generation in 3D	131	73	University of Amsterdam	Netherlands	Academia
Swin UNETR: Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images	6	73	NVIDIA	USA	Industry
GPT-NeoX-20B: An Open-Source Autoregressive Language Model	50	72	EleutherAI	–	Industry
Online reinforcement learning multiplayer non-zero sum games of continuous-time Markov jump linear systems		72	Anhui University	China	Academia
Detecting Twenty-thousand Classes using Image-level Supervision	35	70	Meta	USA	Industry
Image fusion in the loop of high-level vision tasks: A semantic-aware real-time infrared and visible image fusion network		68	Wuhan University	China	Academia
LAION-5B: An open large-scale dataset for training next generation image-text models	53	66	LAION	Germany	Industry
Denoising Diffusion Restoration Models	0	65	Technion	Israel	Academia
VQGAN-CLIP: Open Domain Image Generation and Editing with Natural Language Guidance	175	64	EleutherAI	–	Industry
CLIP-NeRF: Text-and-Image Driven Manipulation of Neural Radiance Fields	33	63	City University of Hong Kong	Hong Kong	Academia
Solving Quantitative Reasoning Problems with Language Models	139	63	Google	USA	Industry
Masked Autoencoders As Spatiotemporal Learners	120	61	Meta	USA	Industry
Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language	499	59	Google	USA	Industry
ViTAEv2: Vision Transformer Advanced by Exploring Inductive Bias for Image Recognition and Beyond	2	59	University of Sydney	Australia	Academia
Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Tasks	178	58	Microsoft	USA	Industry
Language-driven Semantic Segmentation	95	57	Cornell University	USA	Academia
Vision-Language Pre-Training with Triple Contrastive Learning	34	56	University of Texas at Arlington	USA	Academia
Deep Reinforcement Learning-Based Path Control and Optimization for Unmanned Ships		55	Tongji University	China	Academia
EquiBind: Geometric Deep Learning for Drug Binding Structure Prediction	208	54	MIT	USA	Academia
Omnivore: A Single Model for Many Visual Modalities	89	54	Meta	USA	Industry
Quantifying Memorization Across Neural Language Models	106	54	Google	USA	Industry
DeepFusion: Lidar-Camera Deep Fusion for Multi-Modal 3D Object Detection	36	53	Johns Hopkins University	USA	Academia
Genetic Algorithm-Based Trajectory Optimization for Digital Twin Robots		53	Wuhan University of Science and Technology	China	Academia
Make-A-Scene: Scene-Based Text-to-Image Generation with Human Priors	280	53	Meta	USA	Industry
Discovering faster matrix multiplication algorithms with reinforcement learning		52	DeepMind	UK	Industry
DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation	221	52	Google, Boston University	USA, USA	Industry, Academia
PETR: Position Embedding Transformation for Multi-View 3D Object Detection	4	52	Megvii	China	Industry
Protein structure predictions to atomic accuracy with AlphaFold		51	DeepMind	UK	Industry
ABAW: Valence-Arousal Estimation, Expression Recognition, Action Unit Detection & Multi-Task Learning Challenges	2	50	Queen Mary University of London	UK	Academia
HumanNeRF: Free-viewpoint Rendering of Moving People from Monocular Video	72	50	University of Washington	USA	Academia
UnifiedSKG: Unifying and Multi-Tasking Structured Knowledge Grounding with Text-to-Text Language Models	38	49	University of Hong Kong	Hong Kong	Academia
A Systematic Evaluation of Large Language Models of Code	61	48	Carnegie Mellon University	USA	Academia
Robust Speech Recognition via Large-Scale Weak Supervision	40	48	OpenAI	USA	Industry
Diffusion Models: A Comprehensive Survey of Methods and Applications	274	47	Peking University	China	Academia
Can language models learn from explanations in context?	113	46	DeepMind	UK	Industry
NELA-GT-2021: A Large Multi-Labelled News Dataset for The Study of Misinformation in News Articles	9	46	Rensselaer Polytechnic Institute	USA	Academia
ActionFormer: Localizing Moments of Actions with Transformers	0	44	Nanjing University, 4Paradigm Inc.	China, China	Academia, Industry
Least-to-Most Prompting Enables Complex Reasoning in Large Language Models		44	Google	USA	Industry
Diffusion-LM Improves Controllable Text Generation	253	43	Stanford University	USA	Academia
Overview of The Shared Task on Homophobia and Transphobia Detection in Social Media Comments	0	41	National University of Ireland Galway	Ireland	Academia
Text and Code Embeddings by Contrastive Pre-Training	23	40	OpenAI	USA	Industry
Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality	125	40	Hugging Face	USA	Industry
BLOOM: A 176B-Parameter Open-Access Multilingual Language Model	325	39	BigScience Team	France	Industry
Red Teaming Language Models with Language Models	40	39	DeepMind, New York University	UK, USA	Industry, Academia
Transformer Memory as a Differentiable Search Index	372	39	Google	USA	Industry
Torsional Diffusion for Molecular Conformer Generation	109	38	MIT	USA	Academia
Unified Contrastive Learning in Image-Text-Label Space	66	37	Microsoft	USA	Industry
Benchmarking Generalization via In-Context Instructions on 1, 600+ Language Tasks	149	36	University of Washington	USA	Academia

方法论

首先在Zeta Alpha平台上收集了每年被引用次数最多的论文，然后手动检查了第一次发表日期（通常是arXiv预印本），以便将论文放置在正确的年份。通过在Semantic Scholar上挖掘高被引用的人工智能论文来补充此列表，因为它具有更广泛的覆盖范围和按引用计数排序的能力。这主要会发现来自高影响力的闭源出版商（例如《自然》、爱思唯尔、斯普林格和其他期刊）的补充论文。然后取每篇论文在Google Scholar上的引用次数作为代表性指标，并按此数字对论文进行排序，得出该年度的前100篇论文。对于这些论文，使用GPT-3提取作者、附属机构以及所在的国家，并手动检查这些结果（如果从出版物上无法清楚地看出国家，则取该机构总部所在国家）。有多个附属机构的论文对每个附属机构计数一次。

版权声明：除特殊说明外，本站所有文章均为字节点击原创内容，采用 BY-NC-SA 知识共享协议。原文链接：https://byteclicks.com/47133.html 转载时请以链接形式标明本文地址。转载本站内容不得用于任何商业目的。本站转载内容版权归原作者所有，文章内容仅代表作者独立观点，不代表字节点击立场。报道中出现的商标、图像版权及专利和其他版权所有的信息属于其合法持有人，只供传递信息之用，非商务用途。如有侵权，请联系 gavin@byteclicks.com。我们将协调给予处理。

2022年被引用次数最多的100篇人工智能论文

文章推荐：

标签